OBJECTIVE: The relation between serum TSH levels and risk for recurrence or thyroid carcinoma-related death in patients with differentiated thyroid carcinoma has only been studied to a limited extent. DESIGN: We conducted a single-center observational study in 366 consecutive patients with differentiated thyroid carcinoma, who had all been treated according to the same protocol for initial therapy and follow-up. Median duration of follow-up was 8.85 yr. METHODS: The relation between summarizing variables of unstimulated serum TSH concentrations (25th, 50th, and 75th percentiles, the percentage of suppressed and unsuppressed TSH values) and risk for recurrence or thyroid carcinoma-related death was analyzed by Cox survival analyses in patients with at least four TSH measurements. RESULTS: In Cox regression analysis, we found a positive association between serum TSH concentrations and risk for thyroid carcinoma-related death and relapse, even in initially cured patients. The median of the individual TSH concentrations was the best indicator for thyroid carcinoma-related death (hazard ratio 2.03; confidence interval 1.22-3.37) and relapse (hazard ratio 1.41; confidence interval 1.03-1.95). A threshold of 2 mU/liter differentiated best between relapse-free survival and thyroid carcinoma-related death or relapse. CONCLUSION: Our study supports current guidelines, which advise to aim at TSH levels in the low normal range in cured low-risk patients, whereas TSH levels should be suppressed in noncured or high-risk patients.